Detecting and Identifying Selection Structure in Sequential Data
Authors: Yujia Zheng, Zeyu Tang, Yiwen Qiu, Bernhard Schölkopf, Kun Zhang
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we empirically validate our proposed theory and algorithm on both synthetic and real-world datasets. |
| Researcher Affiliation | Academia | 1Carnegie Mellon University 2Max Planck Institute for Intelligent Systems 3MBZUAI. |
| Pseudocode | Yes | Algorithm 1: Identification of the Selection Structure (Sketch with full details in Appendix A). |
| Open Source Code | No | The implementation of tests is based on causal-learn (Zheng et al., 2024). |
| Open Datasets | No | The synthetic datasets are generated according to the selection process on the data generated by linear Gaussian Structural Causal Models (SCMs). In our experiments, we consider guitar soundtrack clips from Hotel California performed by Eagles, and string section soundtrack clips from Panis Angelicus, a verse from the hymn Sacris solemniis and set to music by C esar Franck. |
| Dataset Splits | No | The paper does not provide specific dataset split information (e.g., train/validation/test percentages or sample counts). |
| Hardware Specification | No | All experiments are from 3 random trials with only CPUs and 12 GB of memory. |
| Software Dependencies | No | The implementation of tests is based on causal-learn (Zheng et al., 2024). |
| Experiment Setup | No | The paper describes some data generation and processing steps like 'We sample 10,000 data points' and 'We use a single node to represent each music segment of a length of one bar', and mentions 'Fast Fourier Transform (FFT)' and 'Mel scale conversion', but does not include specific hyperparameters or system-level training settings typically found in experimental setups for machine learning models. |